Hello Ox-users,
I would like to ask a question about an application in [log in to unmask] I am currently
working on a paper , which is simply an out-of-sample forecast comparison
exercise. The data set is a typical FX one.
Estimations are updated each-period, instead of using a fixed set of
coefficients. There is a loop in the program to do that. Naturally there are
convergence problems in some cases (but not more than a handful), especially
with the egarch and fiegarch models. What I would like to do is to detect
these non-convergence problems. I thought that if I put a variable in front
of the doestimation command to capture the return value {e.g.
x=doestimation(.)}, I would get a zero for non-convergence cases. Yet, I am
getting zero in all cases, even when there is strong convergence. My
question is: is there a way to detect and record non-convergence cases?
Moreover, I wonder whether there would be way to use last estimation's
coefficients when there is no convergence. I failed to find a way to impose
old coefficients after I start a new estimation.
I would appreciate your help.
Thanks
Turgut Kisinbay
Economist
Monetary and Forex Operations Division
Monetary and Financial Systems Department
International Monetary Fund
Tel: (202) 623-4158
Fax: (202) 623-6509
_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today - it's FREE!
http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/
|